HIGH-SPEED CAMERA TESTING: EQUIPMENT AND EXPERIMENTAL SETUP Two impact tests were performed on the BU that were designed to target different modes of the structure. In both cases, the cameras were positioned approximately 2 meters from the BU so the entire upright could be seen in the field of view. A 3 lb modal hammer was used to impact the structure. Time data for the hammer was captured using separate, synchronized data acquisition systems. Digital signal processing was carried out in MATLAB and curvefitting was performed using LMS PolyMAX [7,8]. In the first test, a pair of 1.3 Mega-pixel cameras measured the response of the BU due to a perpendicular 1500 lbf impact near Point 1 (see Figure 2a), at one of the top corners of the upright. In this case, only the out-of-plane modes of the upright were excited. The frame rate was set to 500 fps, corresponding to a Nyquist frequency of 250Hz. Three-dimensional pointtracking data was calculated at 30 points evenly distributed on the upright of the BU, including the 8 measurement points common to the other tests. Two averages were taken. For the second impact modal test, the BU was impacted with approximately 4000 lbf at the same point where the shaker input was located for the accelerometer and LDV tests, to excite both the in-plane and out-of-plane modes. Images were taken at a rate of 250 fps with a pair of 3.6 Mega-pixel cameras. The goal was to acquire the first two in-plane modes of the BU which were not measurable from the previous impact test. For this test the Nyquist frequency was 125 Hz and therefore sufficient to capture the modes of interest. A total of 4 averages were taken and the response was tracked with 40 evenly-distributed photogrammetric targets distributed on the structure. EXPERIMENTAL CONSIDERATIONS The experimental considerations in the two impact tests were very similar to those of any other impact modal survey, but with a few additional things requiring attention. As with any impact test, the consistency of the input was a concern because of the variations in the force input due to the human excitation. While the level of amplitude varied somewhat, the input spectra were monitored to ensure a fairly uniform amount of energy was distributed across the frequency ranges of interest. Also, force/exponential windows were applied as needed to reduce the effects of leakage. Beyond these standard concerns, the main issue was the synchronous triggering of the cameras with the other data acquisition systems. Typically, a DAQ system is triggered from the force signal generated by the impact hammer with a pre-trigger delay that captures the beginning part of the transient that would have otherwise been lost. Modern high-end, high-speed cameras have pre-trigger capabilities, but these were not available during the first round of testing. Both systems were triggered via an external source prior to impact, so the timing of the impact was not consistent. Due to the timing variation, different windows had to be applied for each average. In the second impact test, the use of a different timing scheme and trigger synchronization mechanism was investigated. Due to a time delay between the two systems, the phase of the input and output spectra were misaligned. As a result, the poles of the frequency response functions were not stable and the FRFs could not be curvefit. Therefore, the linear spectra calculated from the displacements measured by the imaging systems were used to approximate mode shapes. This procedure provided useful data that were used in correlation studies and for comparisons with the results from the finite element model and with tri-axial accelerometers. EXPERIMENTAL RESULTS AND CORRELATION TO THE FEM The initial results obtained from the first impact test (out-of-plane impact, 500 fps) showed a very high level of correlation to the reference finite element model when compared using the Modal Assurance Criterion (MAC) [9,10] and PseudoOrthogonality Check (POC) [11,12]. Tables 1 through 3 summarize the correlation results for the first impact test. For modes 1 (26 Hz), 3 (78 Hz), and 5 (158 Hz), the diagonal MAC values are 99.9+, 99.8, and 98.1 percent, respectively. The average frequency difference is -0.14%. The diagonal POC terms are all within 3.2% of correlation to the FEA and the offdiagonal terms are all less than 2%. 246
RkJQdWJsaXNoZXIy MTMzNzEzMQ==